Share

Design embedded systems optimized for form factor, power, and cost

CEA-List’s embedded systems design platform develops frugal, compact, low-power, high-performance digital solutions. These solutions, built using increasingly heterogeneous embedded computing resources, are both reliable and secure. The platform is home to hardware and software that can be used to explore, optimize, and implement solutions that meet the needs of highly-constrained embedded systems.

Overview

Unparalleled expertise in hardware-software integration

CEA-List’s embedded systems design platform leverages the institute’s unparalleled expertise in hardware and software design to develop compact, low-power, high-performance digital solutions capable of taking full advantage of increasingly heterogeneous embedded computing resources. For certain applications, these solutions must also meet specific reliability and/or security requirements.

Research at the platform mainly addresses mobility, energy, health, defense, and IIoT use cases, with the goal of delivering operational Edge solutions for situations that require moving computing as close as possible to the sensors that produce the data. The Edge computing solutions developed support the effective implementation of AI systems.

In addition to CEA-List’s extensive know-how in embedded systems design, the platform also has specific resources for designing mixed-criticality systems that are safe and very secure. These resources include advanced prototyping capabilities for multi-sensor systems, embedded AI design, and sensor data fusion (from visible image, infrared, LiDAR, radar, and other sensors). The platform can design and integrate complete systems as well as subsystems of existing systems.

High-level simulation tools developed at CEA-List can be used to explore, evaluate, and validate systems in their operating environments. CEA-List’s holistic approach, which connects design to all requirements for the target application, has earned the trust of companies of all sizes, from major corporations (ArcelorMittal, Renault, Safran, STMicroelectronics, Valeo, etc.), to startups and SMEs.

100

patents, with 12 new patents filed each year

€2 M

in investments

100

employees

Embedded systems exploration and design

The platform provides its R&D partners with all of the tools needed for even the most demanding projects:

  • System exploration and validation with the SESAM integrated systems design environment. SESAM is used for rapid design space exploration, simulation, and validation of complex systems, providing a wide range of processor and peripheral models. It can also model complex memory hierarchies and estimate their performance.
  • Formal analysis of the safety of critical and autonomous systems and cyberphysical systems (for autonomous vehicles, drones, mobile robots, IoT devices, etc).
  • Embedded AI application design, with N2D2, a tool used to explore and generate optimized deep neural networks. The software’s benchmarking features can automatically evaluate the best hardware target for a neural network.

Development of cognitive perception systems

The platform possesses technology bricks, which span hardware, software, and algorithms, for the development of embedded environmental perception systems. Multi-mode systems that are reliable, efficient, and frugal in terms of power, computing resources, and memory, can be explored and implemented using these technology bricks. Most of the platform’s development work leverages three main technologies:

  • ThermalFusion, an algorithmic, software, and hardware solution for the development of mobile environmental perception systems combining visible and infrared vision.
  • SigmaFusion, a frugal algorithmic software solution for Bayesian multi-sensor data fusion (LiDAR, radar, and imagers for 2D and 3D occupancy grids) to be used in conjunction with neural network solutions.
  • LEN-Safety, a frugal software solution that allows autonomous mobile devices to navigate safely without an accurate map of the environment.

Intelligent, secure orchestration of time constrained systems

The platform offers computational and communication models and implementation of the models in software. The resulting integrated monitors ensure operating safety and cybersecurity “by construction” and continuity between Edge and cloud computing for time constrained complex systems. These solutions can be deployed on any system with distributed service level agreement requirements.

An environmental perception solution optimized for embedded systems

This miniature robot-car is equipped with a LiDAR and a camera. It leverages the SigmaFusion sensor data fusion environment to generate 3D point clouds from the data collected. SigmaFusion, which is reliable and frugal in terms of computing resources and power consumption, meets the requirements of highly constrained embedded systems.

Circuit board test bench

This test bench is used to evaluate how well ultrasound recognizes movements with the Esperanto pulse neural network chip.

Sophisticated obstacle-detection systems like the ones found in autonomous vehicles are usually big and heavy and have high power and computing budgets. The INSPEX team will be working on miniaturizing the technology and adapting it for individual and personal applications, which will be even more demanding in terms of detecting obstacles in all conditions.

Suzanne Lesecq

Inspex project coordinator — CEA-Leti
Success story

Inspex packs in innovations for the visually impaired

Inspex is a compact unit that is mounted on a white cane to help the visually impaired user move around safely. An integrated environmental perception system alerts the user to nearby obstacles so that collisions (and even contact with the cane) can be avoided.

The underlying technology is the same as that found in autonomous vehicles. However, it did have to be made much, much smaller and optimized to meet the size, weight, and power requirements of this application.

Inspex was developed under the EU H2020 program with nine European partners including startup GoSense. It won a Étoiles de l’Europe (Stars of Europe) award.

Watch the interview with the project coordinator at the Etoiles de l’Europe 2020 awards ceremony on YouTube.

Listen to the podcast "Parole d'experts, avec Fabien Clermidy, l'Intelligence artificielle embarquée"

Contact

  • Magaly GOUTTEBROZE

    CEA de Saclay / Nano-INNOV
    Bâtiment 862
    Point Courrier 172
    91191 GIF SUR YVETTE Cédex

À lire également

Software development environments

SESAM

SESAM, for faster design space exploration.
Read more
Software development environments

N2D2

N2D2 is used to optimize neural networks, embed them on components or on dedicated hardware accelerators, and measure their performance.
Read more